| commit | 22c350103dc2be48fdd45ea11b077c5d45dd0e15 | |
| author | Brad Werth <bwerth@mozilla.com> | |
| Thu, 13 Jan 2022 19:42:37 +0000 (13 19:42 +0000) | ||
| committer | Brad Werth <bwerth@mozilla.com> | |
| Thu, 13 Jan 2022 19:42:37 +0000 (13 19:42 +0000) | ||
| tree | 6d2a4dd6250762245d48035daf7114410e80a2c7 | treesnapshot (tar.gz zip) |
| parent | e4407de1adb2be5a1f88c49cf783de9a1da3ed29 | commitdiff |
Bug 1738926 Part 2: Stop enqueueing video when the layer gets overwhelmed. r=gfx-reviewers,mstange
There's no point in preparing a video sample when the layer can't accept it.
Limiting the enqueing in this case seems to prevent the macOS WindowServer
from maxxing out the CPU. Instead, the frame will get displayed using the
non-video method of directly setting the layer contents. The user will not
experience an interruption, but the layer will *never* hit detached mode
again. That's an acceptable outcome given that they layer was already
failing to handle the incoming frames and so detached mode was not
being achieved anyway.
Differential Revision: https://phabricator.services.mozilla.com/D135536
There's no point in preparing a video sample when the layer can't accept it.
Limiting the enqueing in this case seems to prevent the macOS WindowServer
from maxxing out the CPU. Instead, the frame will get displayed using the
non-video method of directly setting the layer contents. The user will not
experience an interruption, but the layer will *never* hit detached mode
again. That's an acceptable outcome given that they layer was already
failing to handle the incoming frames and so detached mode was not
being achieved anyway.
Differential Revision: https://phabricator.services.mozilla.com/D135536
| gfx/layers/NativeLayerCA.mm | diffblobblamehistory |